blob: dd287b82f21857a9c04ccbf52314cd879155a5ee [file] [log] [blame]
/*
* Copyright 2018 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "include/gpu/GrDirectContext.h"
#include "include/core/SkTraceMemoryDump.h"
#include "include/gpu/GrContextThreadSafeProxy.h"
#include "src/core/SkAutoMalloc.h"
#include "src/core/SkTaskGroup.h"
#include "src/core/SkTraceEvent.h"
#include "src/gpu/GrBackendUtils.h"
#include "src/gpu/GrClientMappedBufferManager.h"
#include "src/gpu/GrContextThreadSafeProxyPriv.h"
#include "src/gpu/GrDirectContextPriv.h"
#include "src/gpu/GrDrawingManager.h"
#include "src/gpu/GrGpu.h"
#include "src/gpu/GrResourceProvider.h"
#include "src/gpu/GrShaderUtils.h"
#include "src/gpu/SurfaceContext.h"
#include "src/gpu/effects/GrSkSLFP.h"
#include "src/gpu/gl/GrGLGpu.h"
#include "src/gpu/mock/GrMockGpu.h"
#include "src/gpu/ops/GrSmallPathAtlasMgr.h"
#include "src/gpu/text/GrAtlasManager.h"
#include "src/gpu/text/GrStrikeCache.h"
#include "src/image/SkImage_GpuBase.h"
#ifdef SK_METAL
#include "include/gpu/mtl/GrMtlBackendContext.h"
#include "src/gpu/mtl/GrMtlTrampoline.h"
#endif
#ifdef SK_VULKAN
#include "src/gpu/vk/GrVkGpu.h"
#endif
#ifdef SK_DIRECT3D
#include "src/gpu/d3d/GrD3DGpu.h"
#endif
#ifdef SK_DAWN
#include "src/gpu/dawn/GrDawnGpu.h"
#endif
#include <memory>
#if GR_TEST_UTILS
# include "include/utils/SkRandom.h"
# if defined(SK_ENABLE_SCOPED_LSAN_SUPPRESSIONS)
# include <sanitizer/lsan_interface.h>
# endif
#endif
#define ASSERT_SINGLE_OWNER GR_ASSERT_SINGLE_OWNER(this->singleOwner())
GrDirectContext::DirectContextID GrDirectContext::DirectContextID::Next() {
static std::atomic<uint32_t> nextID{1};
uint32_t id;
do {
id = nextID.fetch_add(1, std::memory_order_relaxed);
} while (id == SK_InvalidUniqueID);
return DirectContextID(id);
}
GrDirectContext::GrDirectContext(GrBackendApi backend, const GrContextOptions& options)
: INHERITED(GrContextThreadSafeProxyPriv::Make(backend, options), false)
, fDirectContextID(DirectContextID::Next()) {
}
GrDirectContext::~GrDirectContext() {
ASSERT_SINGLE_OWNER
// this if-test protects against the case where the context is being destroyed
// before having been fully created
if (fGpu) {
this->flushAndSubmit();
}
// We need to make sure all work is finished on the gpu before we start releasing resources.
this->syncAllOutstandingGpuWork(/*shouldExecuteWhileAbandoned=*/false);
this->destroyDrawingManager();
// Ideally we could just let the ptr drop, but resource cache queries this ptr in releaseAll.
if (fResourceCache) {
fResourceCache->releaseAll();
}
// This has to be after GrResourceCache::releaseAll so that other threads that are holding
// async pixel result don't try to destroy buffers off thread.
fMappedBufferManager.reset();
}
sk_sp<GrContextThreadSafeProxy> GrDirectContext::threadSafeProxy() {
return INHERITED::threadSafeProxy();
}
void GrDirectContext::resetGLTextureBindings() {
if (this->abandoned() || this->backend() != GrBackendApi::kOpenGL) {
return;
}
fGpu->resetTextureBindings();
}
void GrDirectContext::resetContext(uint32_t state) {
ASSERT_SINGLE_OWNER
fGpu->markContextDirty(state);
}
void GrDirectContext::abandonContext() {
if (INHERITED::abandoned()) {
return;
}
INHERITED::abandonContext();
// We need to make sure all work is finished on the gpu before we start releasing resources.
this->syncAllOutstandingGpuWork(this->caps()->mustSyncGpuDuringAbandon());
fStrikeCache->freeAll();
fMappedBufferManager->abandon();
fResourceProvider->abandon();
// abandon first so destructors don't try to free the resources in the API.
fResourceCache->abandonAll();
fGpu->disconnect(GrGpu::DisconnectType::kAbandon);
// Must be after GrResourceCache::abandonAll().
fMappedBufferManager.reset();
if (fSmallPathAtlasMgr) {
fSmallPathAtlasMgr->reset();
}
fAtlasManager->freeAll();
}
bool GrDirectContext::abandoned() {
if (INHERITED::abandoned()) {
return true;
}
if (fGpu && fGpu->isDeviceLost()) {
this->abandonContext();
return true;
}
return false;
}
bool GrDirectContext::oomed() { return fGpu ? fGpu->checkAndResetOOMed() : false; }
void GrDirectContext::releaseResourcesAndAbandonContext() {
if (INHERITED::abandoned()) {
return;
}
INHERITED::abandonContext();
// We need to make sure all work is finished on the gpu before we start releasing resources.
this->syncAllOutstandingGpuWork(/*shouldExecuteWhileAbandoned=*/true);
fResourceProvider->abandon();
// Release all resources in the backend 3D API.
fResourceCache->releaseAll();
// Must be after GrResourceCache::releaseAll().
fMappedBufferManager.reset();
fGpu->disconnect(GrGpu::DisconnectType::kCleanup);
if (fSmallPathAtlasMgr) {
fSmallPathAtlasMgr->reset();
}
fAtlasManager->freeAll();
}
void GrDirectContext::freeGpuResources() {
ASSERT_SINGLE_OWNER
if (this->abandoned()) {
return;
}
this->flushAndSubmit();
if (fSmallPathAtlasMgr) {
fSmallPathAtlasMgr->reset();
}
fAtlasManager->freeAll();
// TODO: the glyph cache doesn't hold any GpuResources so this call should not be needed here.
// Some slack in the GrTextBlob's implementation requires it though. That could be fixed.
fStrikeCache->freeAll();
this->drawingManager()->freeGpuResources();
fResourceCache->purgeUnlockedResources();
}
bool GrDirectContext::init() {
ASSERT_SINGLE_OWNER
if (!fGpu) {
return false;
}
fThreadSafeProxy->priv().init(fGpu->refCaps(), fGpu->refPipelineBuilder());
if (!INHERITED::init()) {
return false;
}
SkASSERT(this->getTextBlobCache());
SkASSERT(this->threadSafeCache());
fStrikeCache = std::make_unique<GrStrikeCache>();
fResourceCache = std::make_unique<GrResourceCache>(this->singleOwner(),
this->directContextID(),
this->contextID());
fResourceCache->setProxyProvider(this->proxyProvider());
fResourceCache->setThreadSafeCache(this->threadSafeCache());
#if GR_TEST_UTILS
if (this->options().fResourceCacheLimitOverride != -1) {
this->setResourceCacheLimit(this->options().fResourceCacheLimitOverride);
}
#endif
fResourceProvider = std::make_unique<GrResourceProvider>(fGpu.get(), fResourceCache.get(),
this->singleOwner());
fMappedBufferManager = std::make_unique<GrClientMappedBufferManager>(this->directContextID());
fDidTestPMConversions = false;
// DDL TODO: we need to think through how the task group & persistent cache
// get passed on to/shared between all the DDLRecorders created with this context.
if (this->options().fExecutor) {
fTaskGroup = std::make_unique<SkTaskGroup>(*this->options().fExecutor);
}
fPersistentCache = this->options().fPersistentCache;
GrDrawOpAtlas::AllowMultitexturing allowMultitexturing;
if (GrContextOptions::Enable::kNo == this->options().fAllowMultipleGlyphCacheTextures ||
// multitexturing supported only if range can represent the index + texcoords fully
!(this->caps()->shaderCaps()->floatIs32Bits() ||
this->caps()->shaderCaps()->integerSupport())) {
allowMultitexturing = GrDrawOpAtlas::AllowMultitexturing::kNo;
} else {
allowMultitexturing = GrDrawOpAtlas::AllowMultitexturing::kYes;
}
GrProxyProvider* proxyProvider = this->priv().proxyProvider();
fAtlasManager = std::make_unique<GrAtlasManager>(proxyProvider,
this->options().fGlyphCacheTextureMaximumBytes,
allowMultitexturing);
this->priv().addOnFlushCallbackObject(fAtlasManager.get());
return true;
}
void GrDirectContext::getResourceCacheUsage(int* resourceCount, size_t* resourceBytes) const {
ASSERT_SINGLE_OWNER
if (resourceCount) {
*resourceCount = fResourceCache->getBudgetedResourceCount();
}
if (resourceBytes) {
*resourceBytes = fResourceCache->getBudgetedResourceBytes();
}
}
size_t GrDirectContext::getResourceCachePurgeableBytes() const {
ASSERT_SINGLE_OWNER
return fResourceCache->getPurgeableBytes();
}
void GrDirectContext::getResourceCacheLimits(int* maxResources, size_t* maxResourceBytes) const {
ASSERT_SINGLE_OWNER
if (maxResources) {
*maxResources = -1;
}
if (maxResourceBytes) {
*maxResourceBytes = this->getResourceCacheLimit();
}
}
size_t GrDirectContext::getResourceCacheLimit() const {
ASSERT_SINGLE_OWNER
return fResourceCache->getMaxResourceBytes();
}
void GrDirectContext::setResourceCacheLimits(int unused, size_t maxResourceBytes) {
ASSERT_SINGLE_OWNER
this->setResourceCacheLimit(maxResourceBytes);
}
void GrDirectContext::setResourceCacheLimit(size_t maxResourceBytes) {
ASSERT_SINGLE_OWNER
fResourceCache->setLimit(maxResourceBytes);
}
void GrDirectContext::purgeUnlockedResources(bool scratchResourcesOnly) {
ASSERT_SINGLE_OWNER
if (this->abandoned()) {
return;
}
fResourceCache->purgeUnlockedResources(scratchResourcesOnly);
fResourceCache->purgeAsNeeded();
// The textBlob Cache doesn't actually hold any GPU resource but this is a convenient
// place to purge stale blobs
this->getTextBlobCache()->purgeStaleBlobs();
fGpu->releaseUnlockedBackendObjects();
}
void GrDirectContext::performDeferredCleanup(std::chrono::milliseconds msNotUsed,
bool scratchResourcesOnly) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
ASSERT_SINGLE_OWNER
if (this->abandoned()) {
return;
}
this->checkAsyncWorkCompletion();
fMappedBufferManager->process();
auto purgeTime = GrStdSteadyClock::now() - msNotUsed;
fResourceCache->purgeAsNeeded();
fResourceCache->purgeResourcesNotUsedSince(purgeTime, scratchResourcesOnly);
// The textBlob Cache doesn't actually hold any GPU resource but this is a convenient
// place to purge stale blobs
this->getTextBlobCache()->purgeStaleBlobs();
}
void GrDirectContext::purgeUnlockedResources(size_t bytesToPurge, bool preferScratchResources) {
ASSERT_SINGLE_OWNER
if (this->abandoned()) {
return;
}
fResourceCache->purgeUnlockedResources(bytesToPurge, preferScratchResources);
}
////////////////////////////////////////////////////////////////////////////////
bool GrDirectContext::wait(int numSemaphores, const GrBackendSemaphore waitSemaphores[],
bool deleteSemaphoresAfterWait) {
if (!fGpu || !fGpu->caps()->semaphoreSupport()) {
return false;
}
GrWrapOwnership ownership =
deleteSemaphoresAfterWait ? kAdopt_GrWrapOwnership : kBorrow_GrWrapOwnership;
for (int i = 0; i < numSemaphores; ++i) {
std::unique_ptr<GrSemaphore> sema = fResourceProvider->wrapBackendSemaphore(
waitSemaphores[i], GrSemaphoreWrapType::kWillWait, ownership);
// If we failed to wrap the semaphore it means the client didn't give us a valid semaphore
// to begin with. Therefore, it is fine to not wait on it.
if (sema) {
fGpu->waitSemaphore(sema.get());
}
}
return true;
}
GrSmallPathAtlasMgr* GrDirectContext::onGetSmallPathAtlasMgr() {
if (!fSmallPathAtlasMgr) {
fSmallPathAtlasMgr = std::make_unique<GrSmallPathAtlasMgr>();
this->priv().addOnFlushCallbackObject(fSmallPathAtlasMgr.get());
}
if (!fSmallPathAtlasMgr->initAtlas(this->proxyProvider(), this->caps())) {
return nullptr;
}
return fSmallPathAtlasMgr.get();
}
////////////////////////////////////////////////////////////////////////////////
GrSemaphoresSubmitted GrDirectContext::flush(const GrFlushInfo& info) {
ASSERT_SINGLE_OWNER
if (this->abandoned()) {
if (info.fFinishedProc) {
info.fFinishedProc(info.fFinishedContext);
}
if (info.fSubmittedProc) {
info.fSubmittedProc(info.fSubmittedContext, false);
}
return GrSemaphoresSubmitted::kNo;
}
return this->drawingManager()->flushSurfaces({}, SkSurface::BackendSurfaceAccess::kNoAccess,
info, nullptr);
}
bool GrDirectContext::submit(bool syncCpu) {
ASSERT_SINGLE_OWNER
if (this->abandoned()) {
return false;
}
if (!fGpu) {
return false;
}
return fGpu->submitToGpu(syncCpu);
}
////////////////////////////////////////////////////////////////////////////////
void GrDirectContext::checkAsyncWorkCompletion() {
if (fGpu) {
fGpu->checkFinishProcs();
}
}
void GrDirectContext::syncAllOutstandingGpuWork(bool shouldExecuteWhileAbandoned) {
if (fGpu && (!this->abandoned() || shouldExecuteWhileAbandoned)) {
fGpu->finishOutstandingGpuWork();
this->checkAsyncWorkCompletion();
}
}
////////////////////////////////////////////////////////////////////////////////
void GrDirectContext::storeVkPipelineCacheData() {
if (fGpu) {
fGpu->storeVkPipelineCacheData();
}
}
////////////////////////////////////////////////////////////////////////////////
bool GrDirectContext::supportsDistanceFieldText() const {
return this->caps()->shaderCaps()->supportsDistanceFieldText();
}
//////////////////////////////////////////////////////////////////////////////
void GrDirectContext::dumpMemoryStatistics(SkTraceMemoryDump* traceMemoryDump) const {
ASSERT_SINGLE_OWNER
fResourceCache->dumpMemoryStatistics(traceMemoryDump);
traceMemoryDump->dumpNumericValue("skia/gr_text_blob_cache", "size", "bytes",
this->getTextBlobCache()->usedBytes());
}
GrBackendTexture GrDirectContext::createBackendTexture(int width, int height,
const GrBackendFormat& backendFormat,
GrMipmapped mipMapped,
GrRenderable renderable,
GrProtected isProtected) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
if (this->abandoned()) {
return GrBackendTexture();
}
return fGpu->createBackendTexture({width, height}, backendFormat, renderable,
mipMapped, isProtected);
}
GrBackendTexture GrDirectContext::createBackendTexture(int width, int height,
SkColorType skColorType,
GrMipmapped mipMapped,
GrRenderable renderable,
GrProtected isProtected) {
if (this->abandoned()) {
return GrBackendTexture();
}
const GrBackendFormat format = this->defaultBackendFormat(skColorType, renderable);
return this->createBackendTexture(width, height, format, mipMapped, renderable, isProtected);
}
static GrBackendTexture create_and_clear_backend_texture(GrDirectContext* dContext,
SkISize dimensions,
const GrBackendFormat& backendFormat,
GrMipmapped mipMapped,
GrRenderable renderable,
GrProtected isProtected,
sk_sp<GrRefCntedCallback> finishedCallback,
std::array<float, 4> color) {
GrGpu* gpu = dContext->priv().getGpu();
GrBackendTexture beTex = gpu->createBackendTexture(dimensions, backendFormat, renderable,
mipMapped, isProtected);
if (!beTex.isValid()) {
return {};
}
if (!dContext->priv().getGpu()->clearBackendTexture(beTex,
std::move(finishedCallback),
color)) {
dContext->deleteBackendTexture(beTex);
return {};
}
return beTex;
}
static bool update_texture_with_pixmaps(GrDirectContext* context,
const SkPixmap src[],
int numLevels,
const GrBackendTexture& backendTexture,
GrSurfaceOrigin textureOrigin,
sk_sp<GrRefCntedCallback> finishedCallback) {
GrColorType ct = SkColorTypeToGrColorType(src[0].colorType());
const GrBackendFormat& format = backendTexture.getBackendFormat();
if (!context->priv().caps()->areColorTypeAndFormatCompatible(ct, format)) {
return false;
}
auto proxy = context->priv().proxyProvider()->wrapBackendTexture(backendTexture,
kBorrow_GrWrapOwnership,
GrWrapCacheable::kNo,
kRW_GrIOType,
std::move(finishedCallback));
if (!proxy) {
return false;
}
GrSwizzle swizzle = context->priv().caps()->getReadSwizzle(format, ct);
GrSurfaceProxyView view(std::move(proxy), textureOrigin, swizzle);
skgpu::SurfaceContext surfaceContext(context, std::move(view), src[0].info().colorInfo());
SkAutoSTArray<15, GrCPixmap> tmpSrc(numLevels);
for (int i = 0; i < numLevels; ++i) {
tmpSrc[i] = src[i];
}
if (!surfaceContext.writePixels(context, tmpSrc.get(), numLevels)) {
return false;
}
GrSurfaceProxy* p = surfaceContext.asSurfaceProxy();
GrFlushInfo info;
context->priv().drawingManager()->flushSurfaces({&p, 1},
SkSurface::BackendSurfaceAccess::kNoAccess,
info,
nullptr);
return true;
}
GrBackendTexture GrDirectContext::createBackendTexture(int width, int height,
const GrBackendFormat& backendFormat,
const SkColor4f& color,
GrMipmapped mipMapped,
GrRenderable renderable,
GrProtected isProtected,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
if (this->abandoned()) {
return {};
}
return create_and_clear_backend_texture(this,
{width, height},
backendFormat,
mipMapped,
renderable,
isProtected,
std::move(finishedCallback),
color.array());
}
GrBackendTexture GrDirectContext::createBackendTexture(int width, int height,
SkColorType skColorType,
const SkColor4f& color,
GrMipmapped mipMapped,
GrRenderable renderable,
GrProtected isProtected,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return {};
}
GrBackendFormat format = this->defaultBackendFormat(skColorType, renderable);
if (!format.isValid()) {
return {};
}
GrColorType grColorType = SkColorTypeToGrColorType(skColorType);
SkColor4f swizzledColor = this->caps()->getWriteSwizzle(format, grColorType).applyTo(color);
return create_and_clear_backend_texture(this,
{width, height},
format,
mipMapped,
renderable,
isProtected,
std::move(finishedCallback),
swizzledColor.array());
}
GrBackendTexture GrDirectContext::createBackendTexture(const SkPixmap srcData[],
int numProvidedLevels,
GrSurfaceOrigin textureOrigin,
GrRenderable renderable,
GrProtected isProtected,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return {};
}
if (!srcData || numProvidedLevels <= 0) {
return {};
}
SkColorType colorType = srcData[0].colorType();
GrMipmapped mipMapped = GrMipmapped::kNo;
if (numProvidedLevels > 1) {
mipMapped = GrMipmapped::kYes;
}
GrBackendFormat backendFormat = this->defaultBackendFormat(colorType, renderable);
GrBackendTexture beTex = this->createBackendTexture(srcData[0].width(),
srcData[0].height(),
backendFormat,
mipMapped,
renderable,
isProtected);
if (!beTex.isValid()) {
return {};
}
if (!update_texture_with_pixmaps(this,
srcData,
numProvidedLevels,
beTex,
textureOrigin,
std::move(finishedCallback))) {
this->deleteBackendTexture(beTex);
return {};
}
return beTex;
}
bool GrDirectContext::updateBackendTexture(const GrBackendTexture& backendTexture,
const SkColor4f& color,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return false;
}
return fGpu->clearBackendTexture(backendTexture, std::move(finishedCallback), color.array());
}
bool GrDirectContext::updateBackendTexture(const GrBackendTexture& backendTexture,
SkColorType skColorType,
const SkColor4f& color,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return false;
}
GrBackendFormat format = backendTexture.getBackendFormat();
GrColorType grColorType = SkColorTypeAndFormatToGrColorType(this->caps(), skColorType, format);
if (!this->caps()->areColorTypeAndFormatCompatible(grColorType, format)) {
return false;
}
GrSwizzle swizzle = this->caps()->getWriteSwizzle(format, grColorType);
SkColor4f swizzledColor = swizzle.applyTo(color);
return fGpu->clearBackendTexture(backendTexture,
std::move(finishedCallback),
swizzledColor.array());
}
bool GrDirectContext::updateBackendTexture(const GrBackendTexture& backendTexture,
const SkPixmap srcData[],
int numLevels,
GrSurfaceOrigin textureOrigin,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return false;
}
if (!srcData || numLevels <= 0) {
return false;
}
// If the texture has MIP levels then we require that the full set is overwritten.
int numExpectedLevels = 1;
if (backendTexture.hasMipmaps()) {
numExpectedLevels = SkMipmap::ComputeLevelCount(backendTexture.width(),
backendTexture.height()) + 1;
}
if (numLevels != numExpectedLevels) {
return false;
}
return update_texture_with_pixmaps(this,
srcData,
numLevels,
backendTexture,
textureOrigin,
std::move(finishedCallback));
}
//////////////////////////////////////////////////////////////////////////////
static GrBackendTexture create_and_update_compressed_backend_texture(
GrDirectContext* dContext,
SkISize dimensions,
const GrBackendFormat& backendFormat,
GrMipmapped mipMapped,
GrProtected isProtected,
sk_sp<GrRefCntedCallback> finishedCallback,
const void* data,
size_t size) {
GrGpu* gpu = dContext->priv().getGpu();
GrBackendTexture beTex = gpu->createCompressedBackendTexture(dimensions, backendFormat,
mipMapped, isProtected);
if (!beTex.isValid()) {
return {};
}
if (!dContext->priv().getGpu()->updateCompressedBackendTexture(
beTex, std::move(finishedCallback), data, size)) {
dContext->deleteBackendTexture(beTex);
return {};
}
return beTex;
}
GrBackendTexture GrDirectContext::createCompressedBackendTexture(
int width, int height,
const GrBackendFormat& backendFormat,
const SkColor4f& color,
GrMipmapped mipmapped,
GrProtected isProtected,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return {};
}
SkImage::CompressionType compression = GrBackendFormatToCompressionType(backendFormat);
if (compression == SkImage::CompressionType::kNone) {
return {};
}
size_t size = SkCompressedDataSize(compression,
{width, height},
nullptr,
mipmapped == GrMipmapped::kYes);
auto storage = std::make_unique<char[]>(size);
GrFillInCompressedData(compression, {width, height}, mipmapped, storage.get(), color);
return create_and_update_compressed_backend_texture(this,
{width, height},
backendFormat,
mipmapped,
isProtected,
std::move(finishedCallback),
storage.get(),
size);
}
GrBackendTexture GrDirectContext::createCompressedBackendTexture(
int width, int height,
SkImage::CompressionType compression,
const SkColor4f& color,
GrMipmapped mipMapped,
GrProtected isProtected,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
GrBackendFormat format = this->compressedBackendFormat(compression);
return this->createCompressedBackendTexture(width, height, format, color,
mipMapped, isProtected, finishedProc,
finishedContext);
}
GrBackendTexture GrDirectContext::createCompressedBackendTexture(
int width, int height,
const GrBackendFormat& backendFormat,
const void* compressedData,
size_t dataSize,
GrMipmapped mipMapped,
GrProtected isProtected,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return {};
}
return create_and_update_compressed_backend_texture(this,
{width, height},
backendFormat,
mipMapped,
isProtected,
std::move(finishedCallback),
compressedData,
dataSize);
}
GrBackendTexture GrDirectContext::createCompressedBackendTexture(
int width, int height,
SkImage::CompressionType compression,
const void* data, size_t dataSize,
GrMipmapped mipMapped,
GrProtected isProtected,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
GrBackendFormat format = this->compressedBackendFormat(compression);
return this->createCompressedBackendTexture(width, height, format, data, dataSize, mipMapped,
isProtected, finishedProc, finishedContext);
}
bool GrDirectContext::updateCompressedBackendTexture(const GrBackendTexture& backendTexture,
const SkColor4f& color,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return false;
}
SkImage::CompressionType compression =
GrBackendFormatToCompressionType(backendTexture.getBackendFormat());
if (compression == SkImage::CompressionType::kNone) {
return {};
}
size_t size = SkCompressedDataSize(compression,
backendTexture.dimensions(),
nullptr,
backendTexture.hasMipmaps());
SkAutoMalloc storage(size);
GrFillInCompressedData(compression,
backendTexture.dimensions(),
backendTexture.mipmapped(),
static_cast<char*>(storage.get()),
color);
return fGpu->updateCompressedBackendTexture(backendTexture,
std::move(finishedCallback),
storage.get(),
size);
}
bool GrDirectContext::updateCompressedBackendTexture(const GrBackendTexture& backendTexture,
const void* compressedData,
size_t dataSize,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto finishedCallback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return false;
}
if (!compressedData) {
return false;
}
return fGpu->updateCompressedBackendTexture(backendTexture,
std::move(finishedCallback),
compressedData,
dataSize);
}
//////////////////////////////////////////////////////////////////////////////
bool GrDirectContext::setBackendTextureState(const GrBackendTexture& backendTexture,
const GrBackendSurfaceMutableState& state,
GrBackendSurfaceMutableState* previousState,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto callback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return false;
}
return fGpu->setBackendTextureState(backendTexture, state, previousState, std::move(callback));
}
bool GrDirectContext::setBackendRenderTargetState(const GrBackendRenderTarget& backendRenderTarget,
const GrBackendSurfaceMutableState& state,
GrBackendSurfaceMutableState* previousState,
GrGpuFinishedProc finishedProc,
GrGpuFinishedContext finishedContext) {
auto callback = GrRefCntedCallback::Make(finishedProc, finishedContext);
if (this->abandoned()) {
return false;
}
return fGpu->setBackendRenderTargetState(backendRenderTarget, state, previousState,
std::move(callback));
}
void GrDirectContext::deleteBackendTexture(GrBackendTexture backendTex) {
TRACE_EVENT0("skia.gpu", TRACE_FUNC);
// For the Vulkan backend we still must destroy the backend texture when the context is
// abandoned.
if ((this->abandoned() && this->backend() != GrBackendApi::kVulkan) || !backendTex.isValid()) {
return;
}
fGpu->deleteBackendTexture(backendTex);
}
//////////////////////////////////////////////////////////////////////////////
bool GrDirectContext::precompileShader(const SkData& key, const SkData& data) {
return fGpu->precompileShader(key, data);
}
#ifdef SK_ENABLE_DUMP_GPU
#include "include/core/SkString.h"
#include "src/utils/SkJSONWriter.h"
SkString GrDirectContext::dump() const {
SkDynamicMemoryWStream stream;
SkJSONWriter writer(&stream, SkJSONWriter::Mode::kPretty);
writer.beginObject();
writer.appendString("backend", GrBackendApiToStr(this->backend()));
writer.appendName("caps");
this->caps()->dumpJSON(&writer);
writer.appendName("gpu");
this->fGpu->dumpJSON(&writer);
writer.appendName("context");
this->dumpJSON(&writer);
// Flush JSON to the memory stream
writer.endObject();
writer.flush();
// Null terminate the JSON data in the memory stream
stream.write8(0);
// Allocate a string big enough to hold all the data, then copy out of the stream
SkString result(stream.bytesWritten());
stream.copyToAndReset(result.writable_str());
return result;
}
#endif
#ifdef SK_GL
/*************************************************************************************************/
sk_sp<GrDirectContext> GrDirectContext::MakeGL(sk_sp<const GrGLInterface> glInterface) {
GrContextOptions defaultOptions;
return MakeGL(std::move(glInterface), defaultOptions);
}
sk_sp<GrDirectContext> GrDirectContext::MakeGL(const GrContextOptions& options) {
return MakeGL(nullptr, options);
}
sk_sp<GrDirectContext> GrDirectContext::MakeGL() {
GrContextOptions defaultOptions;
return MakeGL(nullptr, defaultOptions);
}
#if GR_TEST_UTILS
GrGLFunction<GrGLGetErrorFn> make_get_error_with_random_oom(GrGLFunction<GrGLGetErrorFn> original) {
// A SkRandom and a GrGLFunction<GrGLGetErrorFn> are too big to be captured by a
// GrGLFunction<GrGLGetError> (surprise, surprise). So we make a context object and
// capture that by pointer. However, GrGLFunction doesn't support calling a destructor
// on the thing it captures. So we leak the context.
struct GetErrorContext {
SkRandom fRandom;
GrGLFunction<GrGLGetErrorFn> fGetError;
};
auto errorContext = new GetErrorContext;
#if defined(SK_ENABLE_SCOPED_LSAN_SUPPRESSIONS)
__lsan_ignore_object(errorContext);
#endif
errorContext->fGetError = original;
return GrGLFunction<GrGLGetErrorFn>([errorContext]() {
GrGLenum error = errorContext->fGetError();
if (error == GR_GL_NO_ERROR && (errorContext->fRandom.nextU() % 300) == 0) {
error = GR_GL_OUT_OF_MEMORY;
}
return error;
});
}
#endif
sk_sp<GrDirectContext> GrDirectContext::MakeGL(sk_sp<const GrGLInterface> glInterface,
const GrContextOptions& options) {
sk_sp<GrDirectContext> direct(new GrDirectContext(GrBackendApi::kOpenGL, options));
#if GR_TEST_UTILS
if (options.fRandomGLOOM) {
auto copy = sk_make_sp<GrGLInterface>(*glInterface);
copy->fFunctions.fGetError =
make_get_error_with_random_oom(glInterface->fFunctions.fGetError);
#if GR_GL_CHECK_ERROR
// Suppress logging GL errors since we'll be synthetically generating them.
copy->suppressErrorLogging();
#endif
glInterface = std::move(copy);
}
#endif
direct->fGpu = GrGLGpu::Make(std::move(glInterface), options, direct.get());
if (!direct->init()) {
return nullptr;
}
return direct;
}
#endif
/*************************************************************************************************/
sk_sp<GrDirectContext> GrDirectContext::MakeMock(const GrMockOptions* mockOptions) {
GrContextOptions defaultOptions;
return MakeMock(mockOptions, defaultOptions);
}
sk_sp<GrDirectContext> GrDirectContext::MakeMock(const GrMockOptions* mockOptions,
const GrContextOptions& options) {
sk_sp<GrDirectContext> direct(new GrDirectContext(GrBackendApi::kMock, options));
direct->fGpu = GrMockGpu::Make(mockOptions, options, direct.get());
if (!direct->init()) {
return nullptr;
}
return direct;
}
#ifdef SK_VULKAN
/*************************************************************************************************/
sk_sp<GrDirectContext> GrDirectContext::MakeVulkan(const GrVkBackendContext& backendContext) {
GrContextOptions defaultOptions;
return MakeVulkan(backendContext, defaultOptions);
}
sk_sp<GrDirectContext> GrDirectContext::MakeVulkan(const GrVkBackendContext& backendContext,
const GrContextOptions& options) {
sk_sp<GrDirectContext> direct(new GrDirectContext(GrBackendApi::kVulkan, options));
direct->fGpu = GrVkGpu::Make(backendContext, options, direct.get());
if (!direct->init()) {
return nullptr;
}
return direct;
}
#endif
#ifdef SK_METAL
/*************************************************************************************************/
sk_sp<GrDirectContext> GrDirectContext::MakeMetal(const GrMtlBackendContext& backendContext) {
GrContextOptions defaultOptions;
return MakeMetal(backendContext, defaultOptions);
}
sk_sp<GrDirectContext> GrDirectContext::MakeMetal(const GrMtlBackendContext& backendContext,
const GrContextOptions& options) {
sk_sp<GrDirectContext> direct(new GrDirectContext(GrBackendApi::kMetal, options));
direct->fGpu = GrMtlTrampoline::MakeGpu(backendContext, options, direct.get());
if (!direct->init()) {
return nullptr;
}
return direct;
}
// deprecated
sk_sp<GrDirectContext> GrDirectContext::MakeMetal(void* device, void* queue) {
GrContextOptions defaultOptions;
return MakeMetal(device, queue, defaultOptions);
}
// deprecated
// remove include/gpu/mtl/GrMtlBackendContext.h, above, when removed
sk_sp<GrDirectContext> GrDirectContext::MakeMetal(void* device, void* queue,
const GrContextOptions& options) {
sk_sp<GrDirectContext> direct(new GrDirectContext(GrBackendApi::kMetal, options));
GrMtlBackendContext backendContext = {};
backendContext.fDevice.reset(device);
backendContext.fQueue.reset(queue);
return GrDirectContext::MakeMetal(backendContext, options);
}
#endif
#ifdef SK_DIRECT3D
/*************************************************************************************************/
sk_sp<GrDirectContext> GrDirectContext::MakeDirect3D(const GrD3DBackendContext& backendContext) {
GrContextOptions defaultOptions;
return MakeDirect3D(backendContext, defaultOptions);
}
sk_sp<GrDirectContext> GrDirectContext::MakeDirect3D(const GrD3DBackendContext& backendContext,
const GrContextOptions& options) {
sk_sp<GrDirectContext> direct(new GrDirectContext(GrBackendApi::kDirect3D, options));
direct->fGpu = GrD3DGpu::Make(backendContext, options, direct.get());
if (!direct->init()) {
return nullptr;
}
return direct;
}
#endif
#ifdef SK_DAWN
/*************************************************************************************************/
sk_sp<GrDirectContext> GrDirectContext::MakeDawn(const wgpu::Device& device) {
GrContextOptions defaultOptions;
return MakeDawn(device, defaultOptions);
}
sk_sp<GrDirectContext> GrDirectContext::MakeDawn(const wgpu::Device& device,
const GrContextOptions& options) {
sk_sp<GrDirectContext> direct(new GrDirectContext(GrBackendApi::kDawn, options));
direct->fGpu = GrDawnGpu::Make(device, options, direct.get());
if (!direct->init()) {
return nullptr;
}
return direct;
}
#endif